Non-contact multi-parameter sensor network and multi-mode sensing method thereof at elbow joint

Abstract

The invention belongs to the field of ring main units, and particularly relates to a non-contact type multi-parameter sensing network and a multi-mode sensing method of the non-contact type multi-parameter sensing network at an elbow-shaped connector. According to the invention, functions of temperature, partial discharge, cavity humidity, explosion characteristic gas, smoke detection and flame detection are integrated, and through non-contact cooperative detection of six characteristic parameters of heat, humidity, electricity, light, gas and flame, comprehensive capture of a multi-physical field coupling effect in a joint insulation degradation process is realized. Compared with a traditional single sensor (for example, only temperature or partial discharge is monitored), composite fault features such as temperature anomaly, humidity permeation, arc discharge, gas precipitation, smoke generation and flame spreading can be synchronously recognized, monitoring blind areas caused by misjudgment of a single parameter are avoided, and the accuracy of fault early warning is remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of ring main units, specifically a non-contact multi-parameter sensor network and its multimodal sensing method in elbow joints. Background Technology

[0002] Currently, the status monitoring of elbow joints in the State Grid distribution network ring main unit relies heavily on single temperature or partial discharge sensors, which suffers from problems such as high false alarm rate, incomplete coverage, and poor long-term stability, making it difficult to support active protection and graded early warning.

[0003] Single temperature probes have the following problems: Digital IC temperature detection has a limited operating range and a slow response speed. It cannot detect temperatures exceeding its own junction temperature. The chip itself needs thermal equilibrium, and the response speed is generally from hundreds of milliseconds to several seconds, which is slow.

[0004] Infrared temperature detection is easily affected by the environment; dust, water vapor, smoke, etc., can absorb infrared radiation, affecting the measurement results. The detection angle is also limited; if it is blocked or the angle is off, detection errors will occur.

[0005] Partial discharge detection methods have the following problems: In complex industrial environments with severe high-voltage electromagnetic interference, partial discharge detection often exhibits significant signal lag, making rapid and accurate fault location difficult. Existing detection methods are mostly limited to specific frequency bands or preset thresholds, resulting in a narrow range of applications. They are also susceptible to interference from various electromagnetic noises in the field, leading to low reliability of detection results and frequent false alarms. Furthermore, such equipment typically has high requirements for installation location and operating environment, with complex deployment processes and high maintenance difficulty, further limiting its reliability and practicality in real-time early warning scenarios, resulting in unsatisfactory overall early warning performance.

[0006] While existing monitoring methods can reflect local conditions to some extent, these technologies have three fundamental problems.

[0007] First, the sensor units are scattered and large, making it impossible to arrange them in the limited space inside the ring main unit's cable compartment. The different installation positions of each sensor element lead to spatial asynchrony.

[0008] Second, the sensor data acquisition cycles are inconsistent, lacking a unified time benchmark, making it impossible to form multidimensional feature relationships within the same spatiotemporal framework.

[0009] Third, existing algorithms primarily rely on single-threshold judgments or independent parameter trend analysis, failing to reveal the causal relationships and evolutionary patterns among multimodal physical processes, resulting in low accuracy in early fault warnings. They cannot effectively detect and differentiate faults, and the data sources are inaccurate, making it impossible to truly and effectively monitor the working status of elbow joints. Furthermore, they cannot provide accurate data for security warning analysis.

[0010] Therefore, the present invention provides a non-contact multi-parameter sensor network and its multimodal sensing method in an elbow joint. Summary of the Invention

[0011] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0012] The technical solution adopted by this invention to solve its technical problem is: the non-contact multi-parameter sensor network of this invention, which includes: Multi-parameter sensing integrated module: Multiple parameter sensing units are deployed in a non-contact manner at the elbow joint, integrating functions such as temperature, partial discharge, cavity humidity, explosive characteristic gas, smoke detection, and flame detection into one, and working together to achieve multi-parameter sensing of heat, humidity, electricity, light, gas, and flame. Signal preprocessing module: used for time-series synchronization and normalization preprocessing of multi-source heterogeneous signals from different sensors such as temperature, humidity, ultraviolet, gas concentration, and smoke; Data calibration module: Introduces a spatiotemporal compensation algorithm to calibrate probe data and avoid cross-interference and environmental drift; Multi-dimensional feature fusion module: It introduces an algorithm that uses multi-dimensional feature weighting fusion to integrate data from multiple types of sensors, including temperature, humidity, ultraviolet light, smoke, and gas, to comprehensively monitor the potential hazards of elbow joints in real time.

[0013] Furthermore, the multi-parameter sensing integration module includes: Environmental Parameter Detection Unit: Employs digital temperature and humidity sensors to monitor the temperature and humidity in the cable room where the elbow joint is located in real time, and directly outputs calibrated digital signals; uses temperature and humidity parameters as compensation references for other detections, and simultaneously detects high temperature and high humidity in the working environment to determine air ionization environmental parameters; Insulation aging parameter detection unit: For the smoldering characteristic gases released from the insulation layer of the high-temperature triggered joint, a metal oxide semiconductor gas sensor is used for detection. The metal oxide semiconductor sensor is sensitive to volatile organic compounds such as alkanes and aromatic hydrocarbons; an electrochemical sensor is used to detect the released CO; and a smoke sensor based on the laser scattering principle is used to detect the smoke generated by the fire. Partial discharge detection unit: It uses ultraviolet photodetector to detect electric arcs, electric sparks and open flames generated by partial discharge. It uses a solar-blind ultraviolet phototube that is sensitive to ultraviolet light in the 240-280nm wavelength band, avoiding interference from visible light, infrared light and ordinary lighting light. It only responds when it detects ultraviolet light from corona discharge, electric arc ionization radiation.

[0014] Furthermore, the spatiotemporal compensation algorithm includes: a. The probe synchronously samples to obtain the original multi-channel data Xraw(t) under the same time reference; b. Preprocessing, such as noise reduction, zero bias correction, and filtering → X(t); c. Apply the spatiotemporal coupling compensation matrix → X'(t) to obtain the standardized value after compensation; d. Feature extraction, such as trend slope, derivative, instantaneous pulse count, and frequency band energy → feature vector Φ(t); e. Time delay / delay estimation, calculate τi between each channel; f. Calculate the weighted time distance D to measure the multivariate dynamic offset; g. Map d to the fault evolution chain and calculate the posterior probability P; h. Output judgment and confidence level, and trigger early warning / alarm when necessary; The multi-parameter original vector obtained by system sampling is:

[0015] in: T(t): heat, temperature, representing the local temperature detected by the probe, in °C; H(t): Humidity, relative humidity, representing the relative humidity of the air in the monitored area, in %RH; E(t): Electricity, representing the intensity of the local electric field or discharge signal, which can be the electric field amplitude V / m or the instantaneous voltage of the electric arc discharge; L(t): Light, representing the intensity of ultraviolet / visible radiation or pulse count, in units of pulse / s or mV; G(t): gas, representing the concentration of insulating decomposition gases such as CO and alkane VOCs or the output value of a gas sensor, in ppm; F(t): Flame, representing the intensity of flame or infrared radiation signal, in mV or radiation power W / m²; Vector X(t) is the original six-mode signal input synchronously acquired by the probe at time t, which is used for subsequent spatiotemporal compensation and fault evolution analysis.

[0016] Furthermore, the compensated vector X'(t) is: ; The temporal memory effect is represented by two parts: an instantaneous multiplier matrix term and a convolutional hysteresis term. ; Where X[n] = X(t = nΔt), M(H,T) is a 6×6 dynamic matrix, K[k] is a 6×6 matrix sequence representing the hysteresis / memory effect, and Nτ is the maximum hysteresis window size; Set M as the baseline identity matrix plus a small temperature and humidity dependent offset matrix: ; Where I6 is a 6th-order identity matrix, J H J T The cross-sensitivity coefficient matrix is ​​6×6, and the constant matrix is ​​obtained through calibration; α(H,H0) and β(T,T0) are scaling functions representing the intensity of the effect of relative humidity / temperature deviating from the reference value, and can be in linear or saturated form. ; Among them, J H J T Element (J) H ) ij Indicates: the influence coefficient of the original channel j on channel i when the humidity changes; K[k] is used to describe the historical impact of airflow delay and sensor response time. Each element of K[k] is obtained through system identification / calibration and approximated using an exponential decay form. ; Where γ and λ are global scaling parameters, J K This is the directional coefficient matrix.

[0017] Furthermore, a multimodal parametric data fusion algorithm is developed: temporal synchronization and normalization preprocessing are performed on multi-parameter data (thermal, humidity, electrical, optical, gas, and flame). CNN is used to extract features, BiLSTM is used to acquire dependent channels, and a channel attention mechanism is combined to weight the features to achieve multi-dimensional feature fusion. The formula is expressed as follows: ; Wherein, is the high-dimensional representation vector after time-series processing, is the weight, and is the fusion of global feature expression. Combining sensor data and the comprehensive weighted judgment of change trends, the abnormal state of the elbow joint is obtained.

[0018] A non-contact multi-parameter sensor network for multimodal sensing in an elbow joint, the sensing method comprising the following steps: S1. Deploy a non-contact multi-parameter sensor network at the elbow joint of the ring main unit cable compartment; S2. Acquire and preprocess multi-source heterogeneous signals; S3. Apply the spatiotemporal compensation matrix algorithm to calibrate the detection data; S4. Feature extraction and weighted fusion are performed using a multimodal parametric data fusion algorithm; S5 outputs the connector status judgment result and triggers a warning / alarm when necessary.

[0019] The beneficial effects of this invention are as follows: 1. The present invention discloses a non-contact multi-parameter sensor network and its multimodal sensing method for elbow joints. By integrating six characteristic parameters—thermal, humidity, electrical, optical, gas, and flame—for non-contact collaborative detection, it achieves comprehensive capture of the multi-physics coupling effects during joint insulation degradation. Compared to traditional single sensors (such as those monitoring only temperature or partial discharge), it can simultaneously identify complex fault characteristics such as abnormal temperature, humidity penetration, arc discharge, gas evolution, smoke generation, and flame spread, avoiding monitoring blind spots caused by misjudgment of a single parameter and significantly improving the accuracy of fault early warning.

[0020] 2. The non-contact multi-parameter sensor network and its multimodal sensing method in an elbow joint described in this invention employs a spatiotemporal coupling compensation matrix (STCM) algorithm to dynamically calibrate multi-source heterogeneous data. By constructing a 6×6 dynamic compensation matrix M and a historical influence matrix K[k], the influence of environmental factors such as temperature and humidity cross-interference (e.g., increased humidity causing VOC sensor baseline drift) and airflow delay effect (e.g., gas diffusion time lag) on ​​the detection data is accurately quantified.

[0021] 3. The non-contact multi-parameter sensor network and its multimodal sensing method in an elbow joint described in this invention are based on a fusion architecture of convolutional neural network (CNN) and bidirectional long short-term memory network (BiLSTM), combined with a channel attention mechanism, to achieve weighted fusion of spatial features (such as the distribution of partial discharge electric field) and temporal features (such as the slope of temperature trend). By dynamically calculating weights through fully connected layers, key parameters (such as the correlation between the ultraviolet pulse generated by the electric arc and the sudden increase in the concentration of characteristic gases) are highlighted. Attached Figure Description

[0022] The invention will now be further described with reference to the accompanying drawings.

[0023] Figure 1 This is a diagram showing the relationship between the specific parameters of multi-parameter sensing and multi-modal perception in this invention and the target object. Figure 2 This is a flowchart of the multimodal data fusion algorithm in this invention. Detailed Implementation

[0024] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0025] like Figure 1 and Figure 2 As shown, the non-contact multi-parameter sensor network and its multimodal sensing method in an elbow joint according to an embodiment of the present invention are characterized in that the non-contact multi-parameter sensor network includes: Multi-parameter sensing integrated module: Multiple parameter sensing units are deployed in a non-contact manner at the elbow joint, integrating functions such as temperature, partial discharge, cavity humidity, explosive characteristic gas, smoke detection, and flame detection into one, and working together to achieve multi-parameter sensing of heat, humidity, electricity, light, gas, and flame. Signal preprocessing module: used for time-series synchronization and normalization preprocessing of multi-source heterogeneous signals from different sensors such as temperature, humidity, ultraviolet, gas concentration, and smoke; Data calibration module: Introduces a spatiotemporal compensation algorithm to calibrate probe data and avoid cross-interference and environmental drift; Multi-dimensional feature fusion module: It introduces an algorithm that uses multi-dimensional feature weighting fusion to integrate data from multiple sensors such as temperature, humidity, ultraviolet, smoke, and gas. It provides comprehensive, blind-spot-free real-time monitoring of the hidden dangers of elbow joints, solving the difficulties of effectively monitoring multiple parameters of insulation degradation and sensing the joint status.

[0026] The multi-parameter sensing integration module includes: Environmental Parameter Detection Unit: Employs a highly sensitive and precise digital temperature and humidity sensor to detect the temperature and humidity in the cable room where the elbow joint is located in real time, and directly outputs the calibrated digital signal, which has extremely high long-term stability and anti-interference capability; it uses temperature and humidity parameters as an accurate compensation reference for other detections, and at the same time detects the high temperature and high humidity of the working environment to determine the air ionization environmental parameters. Insulation aging parameter detection unit: For smoldering characteristic gases released from the insulation layer of high-temperature triggered joints, a high-precision metal oxide semiconductor (MOS) gas sensor is used for detection. MOS sensors are sensitive to volatile organic compounds (VOCs) such as alkanes and aromatic hydrocarbons. An electrochemical sensor is used to detect the released CO. The electrochemical sensor has high specificity for CO and extremely high sensitivity, and can detect concentration changes at the ppm level. A smoke sensor based on the laser scattering principle is used to detect the smoke generated by combustion. This smoke sensor is more stable than traditional ionization smoke sensors, can distinguish between smoke particles and water vapor, and has an extremely low false alarm rate. Partial discharge detection unit: Utilizes highly reliable military-grade ultraviolet photoelectric detection technology to detect arcs, sparks, and open flames generated by partial discharge. It selects a solar-blind ultraviolet phototube that is sensitive to ultraviolet light in the 240-280nm wavelength band, effectively avoiding interference from visible light, infrared light, and everyday lighting. It only responds when it detects ultraviolet radiation from corona discharge, electric arcs, or other ionizing radiation.

[0027] The spatiotemporal compensation algorithm includes: a. The probe synchronously samples to obtain the original multi-channel data Xraw(t) under the same time reference; b. Preprocessing, such as noise reduction, zero bias correction, and filtering → X(t); c. Apply the spatiotemporal coupling compensation matrix (STCM) → X'(t) to obtain the standardized value after compensation; d. Feature extraction, such as trend slope, derivative, instantaneous pulse count, and frequency band energy → feature vector Φ(t); e. Time delay / delay estimation, calculate τi between each channel; f. Calculate the weighted time distance D to measure the multivariate dynamic offset; g. Map d to the fault evolution chain and calculate the posterior probability P; h. Output judgment and confidence level, and trigger early warning / alarm when necessary; The multi-parameter original vector obtained by system sampling is:

[0028]

[0029] in: T(t): heat, temperature, representing the local temperature detected by the probe, in °C; H(t): Humidity, relative humidity, representing the relative humidity of the air in the monitored area, in %RH; E(t): Electricity, representing the intensity of the local electric field or discharge signal, which can be the electric field amplitude V / m or the instantaneous voltage of the electric arc discharge; L(t): Light, representing the intensity of ultraviolet / visible radiation or pulse count, in units of pulse / s or mV; G(t): gas, representing the concentration of insulating decomposition gases such as CO and alkane VOCs or the output value of a gas sensor, in ppm; F(t): Flame, representing the intensity of flame or infrared radiation signal, in mV or radiation power W / m²; Vector X(t) is the original six-mode signal input synchronously acquired by the probe at time t, which is used for subsequent spatiotemporal compensation and fault evolution analysis.

[0030] The compensated vector X'(t) is: ; The temporal memory effect is represented by two parts: an instantaneous multiplier matrix term and a convolutional hysteresis term. ; Where X[n] = X(t = nΔt), M(H,T) is a 6×6 dynamic matrix, K[k] is a 6×6 matrix sequence representing the hysteresis / memory effect, and Nτ is the maximum hysteresis window size; Set M as the baseline identity matrix plus a small temperature and humidity dependent offset matrix: ; Where I6 is a 6th-order identity matrix, J H J TThe cross-sensitivity coefficient matrix is ​​6×6, and the constant matrix is ​​obtained through calibration; α(H,H0) and β(T,T0) are scaling functions representing the intensity of the effect of relative humidity / temperature deviating from the reference value, and can be in linear or saturated form. ; Among them, J H J T Element (J) H ) ij Indicates: When humidity changes, the influence coefficient (cross sensitivity) of the original channel j on channel i. For example, an increase in humidity will cause the VOC sensor baseline to drift, so the corresponding negative / positive values ​​are filled in the corresponding row and column of M; K[k] is used to describe the historical effects of airflow delay and sensor response time, such as the delay caused by the sample in the microfluidic cavity lingering before entering different sensors. Each element of K[k] is obtained through system identification / calibration and approximated using an exponential decay form: ; Where γ and λ are global scaling parameters, J K This is the directional coefficient matrix.

[0031] Multimodal parametric data fusion algorithm: This algorithm performs time-series synchronization and normalization preprocessing on multi-parameter data (thermal, humidity, electrical, optical, gas, and flame). It employs CNN to extract features, BiLSTM to acquire dependent channels, and combines a channel attention mechanism to weight the features to achieve multi-dimensional feature fusion. The formula is expressed as follows: ; Wherein, is the high-dimensional representation vector after time-series processing, is the weight, and is the fusion of global feature expression. Combining sensor data and the comprehensive weighted judgment of change trends, the abnormal state of the elbow joint is obtained.

[0032] A non-contact multi-parameter sensor network for multimodal sensing in an elbow joint, the sensing method comprising the following steps: S1. Deploy a non-contact multi-parameter sensor network at the elbow joint of the ring main unit cable compartment; S2. Acquire and preprocess multi-source heterogeneous signals; S3. Apply the spatiotemporal compensation matrix algorithm to calibrate the detection data; S4. Feature extraction and weighted fusion are performed using a multimodal parametric data fusion algorithm; S5 outputs the connector status judgment result and triggers a warning / alarm when necessary.

[0033] The multi-parameter, multi-modal sensing method relies on a fusion detector that integrates temperature and humidity, characteristic gases, smoke, and ultraviolet phototubes (detecting partial discharge arcs and flames). It is deployed non-invasively in the cable room of the ring main unit, without changing the line layout or disrupting the spatial magnetic field, and performs multi-parameter data detection in a non-contact manner.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A non-contact multi-parameter sensor network, characterized in that, This non-contact multi-parameter sensor network includes: Multi-parameter sensing integrated module: Multiple parameter sensing units are deployed in a non-contact manner at the elbow joint, integrating functions such as temperature, partial discharge, cavity humidity, explosive characteristic gas, smoke detection, and flame detection into one, and working together to achieve multi-parameter sensing of heat, humidity, electricity, light, gas, and flame. Signal preprocessing module: used for time-series synchronization and normalization preprocessing of multi-source heterogeneous signals from different sensors such as temperature, humidity, ultraviolet, gas concentration, and smoke; Data calibration module: Introduces a spatiotemporal compensation algorithm to calibrate probe data and avoid cross-interference and environmental drift; Multi-dimensional feature fusion module: It introduces an algorithm that uses multi-dimensional feature weighting fusion to integrate data from multiple types of sensors, including temperature, humidity, ultraviolet light, smoke, and gas, to comprehensively monitor the potential hazards of elbow joints in real time.

2. The non-contact multi-parameter sensor network according to claim 1, characterized in that: The multi-parameter sensing integration module includes: Environmental Parameter Detection Unit: Employs digital temperature and humidity sensors to monitor the temperature and humidity in the cable room where the elbow joint is located in real time, and directly outputs calibrated digital signals; uses temperature and humidity parameters as compensation references for other detections, and simultaneously detects high temperature and high humidity in the working environment to determine air ionization environmental parameters; Insulation aging parameter detection unit: For the smoldering characteristic gases released from the insulation layer of the high-temperature triggered joint, a metal oxide semiconductor gas sensor is used for detection. The metal oxide semiconductor sensor is sensitive to volatile organic compounds such as alkanes and aromatic hydrocarbons; an electrochemical sensor is used to detect the released CO; and a smoke sensor based on the laser scattering principle is used to detect the smoke generated by the fire. Partial discharge detection unit: It uses ultraviolet photodetector to detect electric arcs, electric sparks and open flames generated by partial discharge. It uses a solar-blind ultraviolet phototube that is sensitive to ultraviolet light in the 240-280nm wavelength band, avoiding interference from visible light, infrared light and ordinary lighting light. It only responds when it detects ultraviolet light from corona discharge, electric arc ionization radiation.

3. The non-contact multi-parameter sensor network according to claim 1, characterized in that: The spatiotemporal compensation algorithm includes: a. The probe synchronously samples to obtain the original multi-channel data Xraw(t) under the same time reference; b. Preprocessing, such as noise reduction, zero bias correction, and filtering → X(t); c. Apply the spatiotemporal coupling compensation matrix → X'(t) to obtain the standardized value after compensation; d. Feature extraction, such as trend slope, derivative, instantaneous pulse count, and frequency band energy → feature vector Φ(t); e. Time delay / delay estimation, calculate τi between each channel; f. Calculate the weighted time distance D to measure the multivariate dynamic offset; g. Map d to the fault evolution chain and calculate the posterior probability P; h. Output judgment and confidence level, and trigger early warning / alarm when necessary; The multi-parameter original vector obtained by system sampling is:

4. Among them: T(t): heat, temperature, representing the local temperature detected by the probe, in °C; H(t): Humidity, relative humidity, representing the relative humidity of the air in the monitored area, in %RH; E(t): Electricity, representing the intensity of the local electric field or discharge signal, which can be the electric field amplitude V / m or the instantaneous voltage of the electric arc discharge; L(t): Light, representing the intensity of ultraviolet / visible radiation or pulse count, in units of pulse / s or mV; G(t): gas, representing the concentration of insulating decomposition gases such as CO and alkane VOCs or the output value of a gas sensor, in ppm; F(t): Flame, representing the intensity of flame or infrared radiation signal, in mV or radiation power W / m²; Vector X(t) is the original six-mode signal input synchronously acquired by the probe at time t, which is used for subsequent spatiotemporal compensation and fault evolution analysis.

5. The non-contact multi-parameter sensor network according to claim 3, characterized in that: The compensated vector X'(t) is: ； The temporal memory effect is represented by two parts: an instantaneous multiplier matrix term and a convolutional hysteresis term. ； Where X[n] = X(t = nΔt), M(H,T) is a 6×6 dynamic matrix, K[k] is a 6×6 matrix sequence representing the hysteresis / memory effect, and Nτ is the maximum hysteresis window size; Set M as the baseline identity matrix plus a small temperature and humidity dependent offset matrix: ； Where I6 is a 6th-order identity matrix, J H J T The cross-sensitivity coefficient matrix is ​​6×6, and the constant matrix is ​​obtained through calibration; α(H,H0) and β(T,T0) are scaling functions representing the intensity of the effect of relative humidity / temperature deviating from the reference value, and can be in linear or saturated form. ； Among them, J H J T Element (J) H ) ij Indicates: the influence coefficient of the original channel j on channel i when the humidity changes; K[k] is used to describe the historical impact of airflow delay and sensor response time. Each element of K[k] is obtained through system identification / calibration and approximated using an exponential decay form. ； Where γ and λ are global scaling parameters, J K This is the directional coefficient matrix.

6. The non-contact multi-parameter sensor network according to claim 4, characterized in that: Multimodal parametric data fusion algorithm: This algorithm performs time-series synchronization and normalization preprocessing on multi-parameter data (thermal, humidity, electrical, optical, gas, and flame). It employs CNN to extract features, BiLSTM to acquire dependent channels, and a channel attention mechanism to weight the features to achieve multi-dimensional feature fusion. The formula is expressed as follows:

7. Wherein, is the high-dimensional representation vector after time series processing, is the weight, and is the fusion of global feature expression. Combined with sensor data and the comprehensive weighted judgment of change trend, the abnormal state of elbow joint is obtained.

8. A multimodal sensing method for a non-contact multi-parameter sensor network in an elbow joint, used in any one of the non-contact multi-parameter sensor networks described in claims 1-5, characterized in that... The sensing method includes the following steps: S1. Deploy a non-contact multi-parameter sensor network at the elbow joint of the ring main unit cable compartment; S2. Acquire and preprocess multi-source heterogeneous signals; S3. Apply the spatiotemporal compensation matrix algorithm to calibrate the detection data; S4. Feature extraction and weighted fusion are performed using a multimodal parametric data fusion algorithm; S5 outputs the connector status judgment result and triggers a warning / alarm when necessary.

Images